doobidoo/mcp-memory-service v10.55.0

v10.55.0 — Entity Extraction, Insight Cards, urllib3 bump

on GitHub

Highlights

Three changes shipped since v10.54.0, two from @filhocf as part of the #732 reasoning roadmap.

Added

• Entity extraction and memory-entity linking (#868, @filhocf): Phase 2 of the #732 reasoning roadmap. Introduces EntityExtractor that detects @mentions, #tags, URLs, and file paths inside memory content. memory_search gains an entity filter parameter for targeted retrieval; memory_graph gains action="extract_entities" to surface entities from a memory. Entity extraction also runs as Step 5 in the maintain consolidation cycle, continuously indexing entities from new memories.

• Insight Cards — automated pattern/trend/gap detection (#869, @filhocf): Phase 3 of the #732 reasoning roadmap. Adds InsightGenerator that analyses the memory corpus and produces three insight types: patterns (recurring knowledge clusters), trends (frequency changes over time), and gaps (under-represented topic areas). Runs as Step 6 in the maintain consolidation cycle. Opt-in via MCP_INSIGHT_CARDS_ENABLED=true (default: false) to avoid performance impact on existing deployments.

Changed

• Bump urllib3 2.6.3 → 2.7.0 (#893): Routine dependency update.

Upgrade

pip install --upgrade mcp-memory-service

Or with uv:

uv pip install --upgrade mcp-memory-service

New environment variable (optional):

MCP_INSIGHT_CARDS_ENABLED=true   # enable Insight Cards (default: false)

Special Thanks

Huge thanks to @filhocf for continuing to push the #732 reasoning roadmap forward — Phases 2 and 3 in a single release cycle is impressive velocity. The entity extraction and insight card infrastructure lays important groundwork for richer agent cognition.

Full changelog: https://github.com/doobidoo/mcp-memory-service/blob/main/CHANGELOG.md